The Data In Emotion

New techniques may make sentiment analysis seem old hat


In many films involving robots, the one thing they could never have was emotion. They could do everything that a human could and even looked like them, but the thing that marked them apart was a lack of emotional intelligence in either identifying it or having it themselves. However, this may be a thing of the past as computers and systems are now finding the ability to detect and measure emotion.

One of the key ways this is being done is through sentiment analysis across social media, with one of the original studies of its use being Golder & Macy's 'Diurnal and Seasonal Mood Vary with Work, Sleep and Day length Across Diverse Cultures' which used Twitter to ascertain levels of happiness in people. Here, they created a list of hundreds of words with either positive or negative connotation, and after collecting over 500 million messages from 2 millions users across the world, could ascertain happiness based on time of day or season. This work, unsurprisingly, found that people were most happy on longer days and at the weekend whilst moods improved as they moved through the working week towards the weekend.

This kind of work, conducted in the early 2010s, is the precursor to the far more complex emotional data that we are currently seeing or that we are likely to be seeing in the future. We currently have a significant amount of sentiment analysis software available which does this exact work, but on a far larger and more complex scale. Golder & Macy's work was a good start, but would have been slightly inaccurate given that several words can have both positive and negative meanings, like 'sick', 'dope' and 'wicked'. New software analyzes not only words, but word patterns, and attempts to contextualize language, although it is still not a perfect solution.

The future of tracking emotion is going to be most useful in marketing and entertainment, with the former currently leading the field in this regard given the abundance of money spent in the area. The ability to track the success of marketing campaigns is at the heart of modern marketing campaigns, so being able to analyze more than just the actions potentially caused by emotions would be an incredibly powerful tool.

With the proliferation of wearables in the years the come, this may become increasingly easier as companies will have the opportunity to track emotions more accurately. Emotions typically have different physiological affects on the body, which will then be tracked through wearable technology. Scott Byrnes-Fraser, Head of UX design at Adaptive Lab claims 'They can monitor heart rate, blood pressure, temperature, location and movements...Based on that information it would be possible calculate the most likely emotion being felt at that time.'

So through monitoring the impact that specific adverts or experiences have on their audience it will become possible for media creators to create campaigns, programs, or films that they know will have an impact. It allows for feedback to become considerably more thorough, going beyond basic elements like ‘I find this funny/scary/good/bad’ to being able to measure the precise reactions moment-by-moment.

These are obviously some way off, but given how an iPhone can currently track how many steps you take, how many flights of stairs you climb and multiple other metrics, emotional tracking on this scale doesn’t seem so far off.

University lecture small

Read next:

How Are Higher Education Institutions Using Analytics?